00i7-9310/84$3.o.J+0.00 :r: 1984 Pqamon Press Ltd.

Flame radiation in gas turbine combustion chambersARTHUR H. LEFEBVREReilly Professor of Combustion Engineering, School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907, U.S.A.(Received

23 February

1984)

brief description of the nature of flame radiation in gas-turbine combustors is followed by a discussion on the methods and models available for estimating nonluminous and luminous emissivity. Since luminous radiation emanates from soot particles in the flame, consideration is given to the processes of soot formation and soot oxidation, and to the influence ofchemical composition on the soot-forming tendencies of fuels. Experimental data on the effects on flame radiation of variations in the combustor operating conditions of pressure, temperature, velocity, and fuel/air ratio are discussed. The important inihrences on flame radiation of fuel properties and fuel spray characteristics are also considered.

Abstract-A

INTRODUCTION

IT IS now well established that in gas turbine combustion chambers a large proportion of the total heat flux to the liner walls is by radiation from thetrame. In the primary combustion zone most of the radiation emanates from soot particles produced in fuel-rich regions of the flame. Soot may be generated in any part of the combustion zone where fuel/air ratios are high and mixing of fuel and air is inadequate, but the main soot-forming region lies inside the fuel spray at the center of the liner. In this region local pockets of fuel and fuel vapor are enveloped in oxygen-deficient burned products at high temperature, thereby creating conditions that are highly conducive to soot formation, At low pressures the presence of soot particles may give rise to a luminous flame, but usually they are too small in size to signifi~ntIy affect the level of radiation [ 11.However, at the high levels ofpressureencountered in modern gas turbines, the soot particles attain sufficient size and concentration to radiate as black bodies in the infrared region. It is under these conditions that radiant heating is most severe and poses serious problems in regard to iiner durability. Despite its considerable practical importance the process of flame radiation in gas turbine combustors has not been subjected to extensive experimental investigation. In fact, the number of reported studies is surprisingly small. This is due partly to the formidable experimental difficulties involved, especially at high combustion pressures, but another reason is that the designer has always been able to offset the heating effects of flame radiation by the injection of sizeable quantities of film-cooling air along the inner surface of the liner wall. On some engines this amounts to over one-third of the total combustor airflow. The resurgence of interest in flame radiation in gas turbine combustors that occurred in the early 1970s originated from the growing body of evidence on the performance penalties associated with film cooling,KM, *7:9-E

such as deterioration oftemperature pattern factor and combustion inefficiency at low power settings. It was also becoming increasingly recognized that filmcooling air is a main contributor to the presence of carbon monoxide and unburned hydrocarbons in the exhaust gases, especiahy at engine idling conditions. Designers are now fully aware of the paramount importance of reducing the amount of air employed in liner-wall cooling to an absolute minimum. The determination of this minimum quantity for any given combustordepends, ofcourse, on a sound knowledge of flame radiation and the manner and extent to which flame radiation is affected by changes in combustor operating conditions, fuel type, and fuel-spray characteristics. At the present time, the main incentives for a better understanding of flame radiation in gas turbine combustors are the uncertainties surrounding the cost and availability of conventional fuel supplies, indicating the need for examination of the effects of possible changes in fuel properties. The most probable changes are expected to be a decrease in hydrogen content together with increases in aromatic compounds and fuel boiling range. Such changes are expected to promote the formation of soot, ieading to an increase in engine exhaust smoke and, more importantly, increased radiative heat transfer from the flame to the liner. The impact of the latter is emphasized by recent studies which show that small increases in liner wall temperaturecan seriously curtail liner life [2]. A better understanding of radiative heat transfer is also needed to help establish realistic models for the prediction of heat-flux distributions within the combustion zone. Such analytical efforts could yield improved liner durability in future designs by prescribing optimum arrangements for the quantity and distribution of film-cooling air. This approach could also lead to significant reductions in the time and cost of liner development. In recent years a number of excellent reviews have

1493

1494

ARTHUR H. LEFEBVRE

NOMENCLATURE c c2

C co2 C HSCF

C C: C/H r ;, IL. L 1 K

P

constant in equation (18) Plancks second constant correction factor for carbon dioxide correction factor for water vapor fuel concentration oxygen concentration soot concentration carbon/hydrogen ratio by mass activation energy volume fraction of soot particles initial beam intensity beam intensity after distance L path length [m] mean beam length [m] extinction coefficient, or absorption coefficient infrared-average optical constants of soot particles total pressure [kPa]

P 4 R T T, TV

partial pressure [kPa] fuel/air ratio by mass radiative flux, or gas constant temperature [K] initial temperature [K] liner wall temperature [K].

Greek symbols gas absorptivity % nonluminous gas emissivity % soot emissivity s, total emissivity ET liner wall emissivity %V emissivity due to CO, EC01 emissivity due to H,O &Hz0 Stefan-Boltzmann constant equivalence ratio : R specific gravity at 293 K.

been published on the subject of flame radiation [3-51. The present work is confined to flame radiation in gas turbine combustors. Methods of estimating nonluminous radiation are discussed together with various analytical models for flame radiation in enclosures, but attention is focused mainly on the factors that govern the total radiative heat transfer to the liner wall, and the impact on radiative heat flux of combustor design features, combustor operating conditions, fuel composition and fuel spray characteristics. As luminous flame radiation is closely linked to soot formation, some consideration is given to the mechanisms of soot formation and soot oxidation, and to the effect of changes in fuel specifications on soot formation and flame radiation in gas turbine combustion chambers.NONLUMINOUS RADIATION

The nature of the nonluminous radiation emitted by hot gases has beendescribed by Gaydon [6],Tourin [7] and Penner [S]. A characteristic of gas radiation is that it occurs in the form of band spectra. Although solids emit heat at all wavelengths throughout the spectrum, the emissions of a gas are essentially discontinuous and consist of a few narrow bands dispersed over the infrared spectrum. For each molecular specie, radiation occurs at a wavelength that corresponds to the vibration frequency of the atoms within the molecule. As there are several possible modes of vibration, radiation is emitted at a number of different wavelengths. The number, widths, and absorptions of the various bands depend on gas composition and on the pressure, temperature, and thickness of the gas vslume.

In a gas turbine combustion chamber the radiating species are typical of those found in all hydrocarbon flames. The products of combustion consist mainly of H,O, CO,, and N,, along with smaller amounts of CO, NO, NO,, O,, H,, and other minor species. In nonluminous flames the banded spectra from H,O and CO, are the most prominent feature at temperatures up to about 3000 K. The strongest emission bands for H,O arecentered at 1.4,1.9,2.7,6.3, and 21 pm. CO, radiates strongly around 2.7, 4.3, and 15 pm. At higher temperatures H,O and CO, are depleted by dissociation, but radiation from diatomic molecules, notably CO, is increased. Although CO can contribute significantly to the emission and attenuation of radiation within flames, its contribution is localized and is of secondary importance in evaluating radiative fluxes. The contributions of SO, and NO, can be neglected because of their low concentrations. Moreover, gases with symmetrical molecules, such as H,, O,, and N,, give no appreciable radiation, even at the highest temperatures. The fixed location of the emission bands means that the bands gain or lose importance as the temperature changes. Another effect of temperature is to fill up the vibration bands with fine emission lines due to rotational transitions which may broaden and overlap due to various broadening mechanisms and thus become significant [9]. Another important mechanism at low temperatures is collision broadening [lo]. Water vapor is pressure-broadened more than CO,,due to the larger spacings between emission lines in H,O. Emissivity also depends on the size and shape of the gas volume, which are usually combined into a single dimension termed the mean beam length. The longer

Flame radiation

in gas turbine combustion chambers

1495

the mean beam length, the higher will be the emissivity of the gas. Thus the main factors which govern the nonluminous radiation of gases are pressure, temperature, composition, and beam length. Calculation of nonluminous radiation The rate of heat transfer by nonluminous from a gas to its enclosure is given by [ 1 l] R = 0.5 a(l+~,)(~~T~-cx~T~)

radiation (1)

in which E, is dependent on the material, temperature, and degree of oxidation of the wall. Approximate mean values of F, at typical liner wall temperatures for Nimonic, stainless steel, and mild steel are 0.7,0.8, and 0.9, respectively. Investigation over a wide range of values has shown [ 121 that to a sufficiently close approximation T EL-- _ T, % 0 Hence, equation 1.5

as

(2)

-

timel [I31 - - -Lechner [IO] -.-Docherty [9]

1200

1600

2000

(1) may be rewritten

2400

R = 0.5 ~(1 +E,)E~T~.~(T~.-T~). The bulk or mean gas temperature, T: obtained as is the sum of the chamber entry temperature, T,, and the temperature rise due to combustion, ATcomb T= T,+AT,,,,. (3)

FIG. 1. Comparison

of models for carbon

dioxide emissivity.

ALIllb may be derived from standard temperature-rise curves. When these curves are used, the appropriate value for the fuel/air ratio is the product of the local fuel/air ratio and the local level of combustion efficiency. Most heat-transfer calculations are carried out at high pressures, for which it is reasonable to assume a combustion efficiency of 100%. However, this does not apply to the primary zone, where the level of combustion efficiency rarely exceeds 90%, even at the highest pressures. Hottels well-known and well-utilized gas emissivity charts for CO, and H,O have long been the traditional means for estimating the total nonluminous radiation from combustion gases [ 131. In these charts, as shown in Figs. 1 and 2, emissivity is plotted against gas temperature for different values of the product of the partial pressure, P, and the beam length, 1, for a total pressure of 100 kPa (i.e. 1 atm). The total pressure here is the sum of all the partial static pressures and should not be confused with the stagnation pressure. The beam length is determined by the size and shape of the gas volume. For most practical purposes it is given to sufficient accuracy [l l] by the expression I = 3.4 volume surface area (4)

total pressure, the partial pressures vary with fuel/air ratio as shown in Fig. 3. Up to the region near stoichiometric, where dissociation intervenes, this variation may be expressed to a close approximation

WI by2.lq P -=___ P 1+ 1.05q for both carbon dioxide and water vapor.

-

-------Hottel ---Leckner

[I33

-001 -

-.-

[IO1 DOCilerty [9]951

----------Forog

For tubular systems the above expression yields values of beam length ranging from 0.60, to 0.9D,, depending on the length/diameter ratio of the liner. For a long annular liner, 1 N 0.9D,. For a typical kerosine of the form C,H,,, for a fixed

1200

1600

2cco

2400

FIG. 2. Comparison

of models for water vapor emissivity.

1496

ARTHUR H

LEFEBVRE

turbine interest Hottels charts [ 131 suggest that AE is independent of gas temperature above 1200 K, where Aa remains constant at 0.054. Although this is true for most practical purposes, As does in fact vary slightly with temperature in this range, its value rising with increase in temperature from 1200 to around 1600 K, after which it declines slowly with further increase in temperature [lS, 171. Thus for combustion pressures below 1MPa( 10atm) the nonluminous emissivity is given by

0

002

004

006

008

010

Fuel/air

ratio

FIG. 3. Variation of partial pressure of carbon dioxide water vapor with fuel/air ratio [12].

and

All the parameters in the above equation can be obtained (sometimes a little extrapolation is called for) directly from Hottels charts [ 131.For high combustion pressures (> 1 MPa) and temperatures (> 1200 K), where Cccl 2: 1.0, CHlo = 1.42, and BE 2: 0.054, equation (11) simplifies to gg = eCozi- 1.42zHzo- 0.077. (12)

More general empirical formulae for estimating pcot and pHl,-, have been provided by Kunitomo and Kodama [ 141 as P,,,/P 0.175 0.07514 + 0.033R = for 4 < 1.1 (6) = 0.092+0.03%2-2.619(1.01-n) x (LOS- I/#)2 for # 2 1.1 (7) P,,,/P 0.259-0.068/~-0.0783~ = for # > 1.05 (8) = 0.126-0.72[1/&+0.887fi1.724 - 1.42(Q-0.805)2]2 for f#~ 1.05 (9) >

Where a large number of emissivity calculations are involved, using Hottels charts to obtain ECo, and &W2o can be tedious and time consuming. To alleviate this problem various approximate formulae for nonluminous emissivity have been proposed, of which the most widely quoted in gas turbine literature is the following due to Reeves [IS] eg = 1 -exp [-290P(qf)0~5T-1~5]. (13)

Unfortunately this expression is valid only for gas pressures below 5 atm. For higher pressures more accurate values are given by the empirical formula [ 161. Eg= (0.15-0.00005T)(2Pq1)~~~~+~~~~~*~~. (14)This expression yields values of&,which agree closely with the corresponding values from the updated charts for +O2r 8Hz0,Ccoz, CHIo, and A&due to Farag [ 151. as described by Sarofim et al. [ 173. It is considered valid for pressures below 4000 kPa (40 atm), temperatures between 1200 and 2400 K, and fuel/air ratios weaker than stoichiometric.

where 4 is the equivalence ratio and R is the specific gravity of the fuel at 293 K. The vafues of zcoz and cttzoas read off Hottels charts must first be multiplied by the pressure-broadening respectively, before being added factors, Ccol and CH20r together to obtain the total nonluminous emissivity [133. The original work by Hottel included total pressure correction factors for both gases, but the range ofpressures covered was far too low to meet the needs of modern gas turbines. More recent work by Farag [15] has yielded correction factors for CO, and H,O at the higher levels of pressure of interest for gas turbines. At these conditions the correction factor for H,O is given by El61 = CJ&c) 1.44 + P[0.00054/T0.5 -0.002~]. (10)

However, inspection of Farags graphs reveals that in the regions ofmain interest the values of Ccoz and CNlo tend to remain fairly constant at around 1 and 1.42, respectively. Since the emission bands of COz and H,O overlap, each gas absorbs radiation from the other. When both are present, the total radiation emitted is less than the sum of that due to each of the two gases acting alone. Thus, after the emissivities of CO, and H,O are obtained separately, a correction for mutual absorption, AE,must be deducted. For the liquid fuels of gas

In#uence ofjiiel chemistry The composition of the combustion products and, in particular the relative proportions of CO, and H,O varies with the carbon/hydrogen ratio of the fuel. According to Sarofim [19] emissivities increase with hydrogen content up to a value of H/(C + H) of around 0.85. For a change from a petroleum-derived distillate fuel [(CH,), or H/(C + H) = 0.671 to a highly aromatic fuel [(CH),, or H/(C + H) = OS] the emissivity would decrease by just a few percent at high combustion pressures. Thus, for the various alternative fuels of interest for gas turbine applications, the influence offuel type on nonluminous radiation is expected to be quite small.Emissiuity models

Hottels emissivity charts are simple to use and are considered accurate in the regions supported by

Flame radiation

in gas turbine

combustion

chambers

1497

experimental data. However, they are not well suited for mixtures of more than two radiating species, and correction factors for spectral overlap when dealing with mixtures of various gases do not always exist. During the last several decades many attempts have been made to alleviate this situation by providing a more general and fundamental spectral basis for gaseous radiation. The results of these extensive theoretical and experimental studies have been reviewed by de Ris [3], Tien and Lee [4], Viskanta [SJ, and Docherty [9], so only their main features are discussed below. In theory it might be possible to predict the positions and strengths ofall the absorption lines in the spectrum for a gas, and this is the objective of the just overlapping line model of Penner and Varanasi [20]. A less fundamental but more practical approach is the socalled narrow-band model which divides the thermal radiating spectrum into many small intervals for each radiating specie. Within each interval the spectral lines are assumed to have random or regularly spaced locations, while their strengths are assumed to follow an exponential or uniform probability distribution [3]. Ludwig and coworkers [21, 221 have provided the spectral constants needed to calculate for each contributing interval the emission and absorption of most of the important gases over pressures and temperatures of interest in combustion systems. Unfortunately, calculations using the narrow-band model require a massive amount ofinput data and large amounts of computer time which makes the model impractical for most engineering applications. Simpler approaches based on analytical approximations to the results of the narrow-band model have been developed for CO,-H,O mixtures by Leckner [lo] and Taylor and Foster [23]. The results of Leckners calculations for CO, and H,O are plotted in Figs. I and 2. They demonstrate close agreement with Hottels charts except at high temperatures and long beam lengths, where Hottels overlap corrections underpredict the emissivity [3]. Foster and Taylors model also agrees well with Hottels empirical correlations but is restricted to low-pressure applications which are of little interest for the gas turbine. Following a model suggested by Sarofim et al. [17], by Farag [I 51 attempted to the radiation of a number of gray gases. This allows the emissivity-~~ relationship for any real gas to be approximated by the weighted sum of a sufficient number ofgray gases. Farags results for CO, and H,O are included in Figs. 1 and 2, respectively. The agreement of the model with the reference emissivities [ 171 is clearly very satisfactory, but it should be noted that the results presented for H,O are confined to temperatures below 1000 K. During the past two decades Edwards and coworkers [24-261 have developed the exponential wide-band model for radiation from combustion gases. This model takes advantage of the fact that infrared radiation from each specie is concentrated into

up to six wide-band spectral regions associated with its principal vibrational transitions. The numerous individual spectral lines within these vibrational bands are associated with different rotational transitions. The method ailows the width of a given band of gas species to be calculated for any mixture at any temperature and pressure. Some of the species overlap in their absorption bands and this must be accommodated to avoid overestimating the total mixture emissivity. Best-fit parameters have been provided for the wideband model by Edwards and Balakrishnan [25] for the important gases over broad ranges of pressure and temperature. The model has been used successfully by Docherty [9] whose results for CO, and H,O appear in Figs. I and 2, respectively. Modak [27] has since revised the wide-band parameters for H,O which improves the models accuracy at temperatures below 1000 K. De Ris [3] has shown that Modaks improved wide-band model agrees closely with Hottels correlations despite its completely independent source of data. fn summary the wide-band model is applicable under widely different physical conditions, and mixtures of CO*, H,O, CO, NO, SOZ, and CH, can easily be handled. Moreover, it is simple to use and appears to have the advantages of accuracy and generality possessed by the more detailed narrow-band made1 for combustion situations [3]. Further information on the wide-band model, and other models of relevance to combustion systems, is contained in the review articles of de Ris [3], Docherty [9], and Tien and Lee [4].Radiativejux models

In gas turbine combustors a significant proportion of the total heat transferred from the combustion gases to the liner is by radiation. Thus accurate assessments of radiant heat flux are an essential prerequisite for the prediction of liner wail temperatures and liner durability. An additional incentive for establishing methods for predicting radiative exchange in gas turbine combustors stems from their potential contribution to the control and reduction of pollutant emissions. Advanced modeling techniques are being developed for predicting the emissions of carbon monoxide, unburned hydrocarbons, and oxides of nitrogen, but these methods require an accurate description of gas temperature distributions in all regions of the combustor. As the local temperature of any small segment within the flame zone is influenced strongly by the amount of radiant heat it exchanges with neighboring segments and with the liner wall, it follows that accurate assessments of pollutant emissions necessitate fairly precise knowledge of radiant heattransfer rates. These considerations have led in recent years to the development of several models for flame radiation in furnaces and combustion chambers. Detailed information on these models is provided by Viskanta [S], and in the following sections consideration is given only to

1498

ARTHUR H. LEFEBVRE

those which appear most relevant to gas turbine combustors. Perhaps the most widely used method for calculating radiative transfer in non-isothermal enclosures is the zone method as developed by Hottel and Cohen [ZS] and Hottel and Sarofim [29]_ Application of the zone method requires the whole gas volume to be divided into a number of smaller volumes, each with uniform properties such as temperature, composition, emissivity and transmissivity. The radiation from each volume to the liner wall can then be calculated, provided their relative geometrical orientation is known. The method requires only the solution of a set of simultaneous algebraic equations ; however, the computer time involved may be so lengthy as to seriously restrict the number of zones into which the total combustion volume is divided. In addition to limiting the accuracy of the method, this could also lead to incompatibility between the coarse grid used for radiative transfer calculations and the finer grids required for associated fluid-mechanics calculations

caThe Monte-Carlo method is also based on the solution of the simultaneous algebraic equations of three-dimensional (3-D) heat transfer, but the calculations are greatly simplified by treating the radiative exchange between gases and/or surfaces in a probabilistic manner. This allows the number of zones into which the combustion volume is subdivided to be greatly increased. The total energy emitted by each zone is divided into several beams each of which is emitted with equal energy in a random direction from a random location in the zone. As each beam travels through the gas its attenuation is described by Beers law. If its energy is not totally absorbed by the combustion gases before it reaches a wall surface, the remaining energy is reflected back into the combustion volume in a random direction. The calculation continues until the beam energy is completely absorbedc301.

The merits and shortcomings of the Monte-Carlo method have been discussed by Viskanta [S]. Its main advantages lie in its flexibility, and its ability to be applied to complex geometries. Absorption and scattering by particles in the flame can be accounted for adequately. Its principal drawback is that it is time consuming to the extent that it oflers no economic advantage over the zonal method. Furthermore, it is incompatible with the numerical techniques used for fluid flow and temperature field calculations because they require different numerical algorithms. In the flux methods the solid angular distribution of radiation intensity is subdivided into several smaller solid angles. Subsequent integration of the equation of radiant energy transfer over each of the smaller solid angles leads to a set of simultaneous partial differential equations for the unknown intensities, which may be solved in parallel with the differential equations representing balances in mass, momentum, energy, and chemical composition [31].

Radiation in an absorbing-emitting gray gas within a cylindrical enclosure is normally represented by a four-flux model of the radiative field, with two fluxes in each of the radial and axial directions. Two-flux methods have been advocated which involve the use of a one-dimensional (1-D) space grid rather than the twodimensional (2-D) grid required for the four-flux representation [31]. This simpler approach reduces computational complexity and allows considerable economies in computer time, but the accuracy is thereby greatly reduced. Spalding [32] has formulated four- and six-flux methods, and extensions of the two-flux method to multi-~ux and nongray emitting-absorbing media are also discussed by SiddallC333. One reported example of the use of the four-flux method demonstrated satisfactory prediction of wall temperatures, although radiation fluxes were underestimated [34]. A six-flux model is currently in use at the Garrett Turbine Engine Company [35]. The reasons for the inaccuracies in flux methods are {a) the radiant flux is divided into too few directions (2, 4, or 6 being small for many applications), and (b) the fluxes in the different directions are unrealistically independent of each other. Another limitation of the flux models is that their extension to general curvilinear coordinates for handling complex geometries is rather cumbersome [35]. The discrete-transfer method proposed by Lockwood and Shah [36] is similar to the Monte-Carlo method in that it is based on the solution of representatively-directed beams of radiation within the combustor. However, in this method the ray directions are not random but are specified in advance, and are solved only between two boundary walls. The method is economical, easy to apply, and has potential for high accuracy. It can be adapted to a wide range of geometries and is designed to be coupled to fluid flow solutions. The method appears to be well suited for gas turbine combustors and one such application has been reported [35].

LUMINOUS

FLAMES

At atmospheric pressure, the soot particles formed in combustion contribute a continuum in the visible spectral range, thereby producing a luminous flame, but usually they are too small in size to radiate appreciable energy in the important infrared region, With increase in pressure, the continuous radiation increases in intensity and the molecular bands become less pronounced. At the high levels of pressure encountered in modern gas turbines, the soot particles can attain sufficient size to radiate as black bodies in the infrared region, and the flame is then characterized by a predominance of continuous radiation. It has been found that when a parallel beam of radiation passes through a system of particles the strength of the beam declines exponentially according

Fiame radiation

in gas turbine

combustion

chambers

1499

to the relationship

[37] I, /I, = exp ( - KL) (15)

where IL is the intensity of the beam after travelling a distance L through the system of particles, I, is the intensity at L = 0, and I( is the extinction coefficient having the dimensions of reciprocal length. For small particles (no/n < 1) scattering of radiation is small in comparison with absorption and K becomes the absorption coefficient. In recent years a subject of debate has been whether or not luminous flames can be regarded as spectrally gray [3]. This is an important issue, since the gray ~sumption facilitates the reporting of llame property data and also greatly simplifies engineering calculations of radiative heat transfer. If emissivity can be assumed equal to absorptivity, then the emissivity of a soot-bearing medium may be defined as E, = (I, -1,)/f,, Substitution from equation (15) leads to the resultE, =

From a practical standpoint, however, the value and importance of equation (18) is that using this simple expression an experimentalist can determine the sootemission coefficient, E204 mass) of naphthalenes or tettalins exhibit sooting properties that are more dependent on the presence ofsuch components than on the hydrogen content. The important intluence of hydrogen content on flame radiation was first demonstrated by Schirmer and coworkers [99,100] who investigated a wide range offuels at combustion pressures up to 1.5 MPa (15 atm). The results oftheir tests on the Phillips 5 cm combustor showed a systematic increase in radiative flux with increase in combustion pressure and decrease in fuel hydrogen content. More recent work by Kuznar et al. [lOI] and Humenik et al. [102] on much larger gas turbine combustors generally confirmed Schirmers results. Moses and Naegeli also used the Phillips 5 cm combustor, as well as aT-63 enginecombustor, to study the effect of fuel composition on flame radiation [103]. They concluded that hydrocarbon molecular structure has at most a minor effect on flame radiation and that hydrogen content or C/H ratio is the main factor. They also noted that fuel viscosity and boiling point distribution have no significant effect on radiation even with end points as high as 675 K, which implies that the soot-forming reactions are gas-phase rather than liquid-phase pyrolysis. In subsequent studies, by suitable blending Naegeli and coworkers [104, 1051 were able to produce fuels of constant hydrogen content and varying aromatics content. These blended fuels allowed them to study the separate effects on game radiation of C/H ratio and percentage aromatics. Their results showed that the dominant soot formation mechanisms in gas turbine engines are dependent on combustor operating conditions. There appears to be a number of competing soot formation and oxidation steps whose relative importance varies with inlet air temperature, reference velocity, and fuel/air ratio. Naegeh et al. also found that under some operating conditions the sooting tendency is affected significantly by polycyclic aromatics, but the C/H ratio is the principal correlating parameter. Figure 4 shows what was typically observed for the effect of polycyclic aromatic ring carbon on flame radiation. It is apparent in this figure that flame radiation has a linear dependence on polycyclic aromatics content.

FIG. 4. Effects of hydrogen content and polycyclic aromatic ring carbon on flame radiation intensity [105].

According to Clark Cl063 an appropriate correlation parameter for these data would be C%H, -O.O%(o/,PA)]. However, Clark has proposed an alternative fitting parameter having a nonlinear dependence on polycyclic aromaticity of the form CO/oH2 WA)1 where 0.1